The Research On The Forecast Of Urban Pavement Performances And Maintenance Decision-making

Countermeasure of maintenance and investment were influenced by pavement performance directly. In order to distribute given maintenance budget optimally in time and space, and determine the best scheme of pavement maintenance, pavement performance in a certain period must be forecasted. Evaluating pavement condition is an important basis for forecast and decision-making, an excellent forecast model is the foundation for reasonable decision-making, whether decision-making model is reasonable or not, it is the key factor of success or failure for the whole system.Because some defects exist in traditional forecast method in such as sample size and precision, pavement performance forecast based on support vector machines is presented in this study. Taking urban network asphalt roadway as the research object, combining with the reality of the urban road in Guangzhou, PCI index was forecasted. All data were prepared according to the format of LIBSVM-2.84 software package,then the forecast value in 2006 was calculated by means of support vector machines for regression. Meanwhile, multivariate linear regression equation was obtained by MATLAB 7.0, the PCI value in 2006 was forecasted by the equation.The two predictive values were compared with the practical data, the results show that the precision of SVM is more precise than the multivariate linear regression. The forecasting accuracy was improved greatly for using SVM in pavement performance forecast. Subsequently, maintenance countermeasures was analysed in the economic field, the maintenance decision-making was divided into three kinds of situations, and the correlating models were built aiming at the facts in Guangzhou。With LINDO software, selected the model under limited capital and optimum effect, the sample calculation is given based on the results of the evaluating pavement condition and the forecast. Three urban road sections from the network in Guangzhou were selected for the example of calculation. The way of model is reasonable and the method is feasible, the model is practical, thus it has a certain practicability. It can provide effective decision-making for the road administration department.Pavement performance was forecasted by support vector machines in this study. The maintenance decision-making models were built aiming at the facts of urban network asphalt roadway in Guangzhou. These can provide the scientific foundation and decision-making support for the road administration department. This research enriches the theory of pavement management system, extends the application field of support vector machines, and has a certain practical and scientific significance.